California Just Made Anthropic Its Favorite Vendor — While Washington Still Calls Them a Threat

June 30, 2026 — Day 161 of our AI crisis timeline.

Yesterday, Governor Gavin Newsom stood up and did something no other US government entity has done since June 12: he publicly endorsed Anthropic, at scale, in writing, with a press release and his name attached.

California has entered into a new partnership with California-based company Anthropic to help state agencies responsibly use artificial intelligence to improve services, expand state workforce training, and continue serving Californians. Governor Gavin Newsom and Anthropic have made a deal that allows California government agencies to use Claude at a discounted price.

California will give all its state agencies — and every city and county in the state — access to Anthropic’s Claude at a 50% discount starting right away.

This is the same company whose flagship models were globally disabled by a federal export control directive eighteen days ago. The same company the Pentagon labeled a “supply-chain risk.” The same company whose Defense Secretary publicly said, “Three months ago, we kicked Anthropic out of our building — forever. Every passing day proves why that was the right move.”

California’s chief information officer was asked directly whether any of that mattered to the negotiation. CIO Chris Given said the federal supply-chain risk designation against Anthropic “just didn’t come up” during contract negotiations.

It just didn’t come up. Two different governments, both American, looking at the exact same company and reaching opposite conclusions about whether it’s a national security threat or a state procurement priority.

What the Deal Actually Contains

Let’s establish the facts before the politics, because the deal itself is substantial.

The Scope

Claude is the first AI productivity tool that will be available to all State agencies though the California Department of Technology’s new Statewide Information Technology Shared Services (SITeS) portal. The agreement makes Claude the first AI productivity tool available statewide through California’s centralized software portal, and comes bundled with free workforce training and on-call technical support from Anthropic engineers.

The agreement also provides the same discounted offer for California’s local governments, including cities and counties.

This isn’t a pilot program. It’s every state agency, every city, every county — the entire procurement apparatus of the world’s fifth-largest economy, defaulting to Claude at half price.

What Was Already Running

California has already deployed Claude at the DMV, Department of Healthcare Services, and CDT/CalOES. There is also Poppy — a state-built AI assistant named after California’s official flower, designed specifically for government workflows, with pre-built queries tailored to common state business needs. Poppy piloted with 2,800+ state employees across 67 departments and is on track for statewide rollout in July 2026.

So this wasn’t a cold start. The SITeS deal formalizes and centralizes what was already spreading organically. California built a working relationship with Claude well before the June 12 crisis, then formalized it into statewide infrastructure right as the federal relationship hit its lowest point.

The Pricing Signal

“A lot of departments are going to switch their usage to this contract, and that’s very much our intent,” Chris Given, California’s chief information officer, told Politico. “When we see that folks are going to be using a tool more, we want to make sure that we, as the state, have negotiated the best possible price for them.”

50% off isn’t a courtesy discount — it’s a volume bet. California is explicitly trying to drive every department onto this contract. That’s the kind of pricing you offer when you need to demonstrate massive adoption numbers, fast, to a market watching whether anyone still trusts you after a government shutdown.

The Pentagon Backstory This Deal Is Answering

To understand why this matters, you need the full arc of Anthropic’s relationship with the federal government — which goes back further than the June 12 export control.

Earlier this year, Anthropic and the U.S. Department of Defense clashed over a contract that would give the government agency permission to deploy Claude for any lawful use. Anthropic sought to explicitly carve out protections that prevent the government from using its technology to surveil Americans or deploy autonomous weapons without human oversight. But Defense Secretary Pete Hegseth refused, and the agency signed a deal with OpenAI instead. The government went as far as to declare Anthropic a “supply-chain risk,” preventing the company from working with any other Pentagon contractors.

The Pentagon labeled Anthropic a supply-chain risk after the company refused to let the military use Claude for mass domestic surveillance or fully autonomous weapons without human oversight.

This is the foundational dispute. Long before Fable 5’s jailbreak, before the export control directive, before “they screwed us” — Anthropic refused to build mass surveillance and autonomous weapons capability for the Pentagon on principle, and got blacklisted for it.

The June 12 export control over the Fable 5 jailbreak happened against this backdrop of an already-adversarial relationship. It’s not a standalone technical dispute — it’s the latest chapter in a yearlong pattern of friction between Anthropic’s safety commitments and federal demands.

The Court Just Intervened

Buried in one of today’s reports is a detail that changes the calculus significantly: a federal judge has since blocked that Pentagon designation.

A federal court blocked the Pentagon’s supply-chain risk designation against Anthropic. This means the legal basis for treating Anthropic as untrustworthy has been judicially challenged and at least partially invalidated — right as California signs its biggest-ever AI procurement deal with the company.

The timing here is not coincidental in feel, even if it’s coincidental in fact. California moved forward with a major Anthropic partnership in the same window that a court ruled the federal government’s hostility toward Anthropic had exceeded its legal authority.

The Politics Newsom Is Playing

This deal cannot be separated from Newsom’s broader positioning, which multiple outlets flagged explicitly.

The political undertone here is hard to miss. Newsom has spent months positioning California as a counterweight to the Trump administration on AI — on regulation, on worker protection, and on which companies government should trust.

Newsom is term-limited and won’t be on the ballot this fall. His AI positioning is less about electoral politics than about legacy — and California’s ambition to write its own technology policy independent of Washington.

The governor, widely expected to run for president in 2028, is positioning California as a state that adopts AI aggressively while insisting on guardrails, a contrast he has drawn explicitly with the Trump administration’s approach.

There’s a 2028 presidential calculus embedded in a procurement announcement about chatbot licensing. Newsom isn’t just buying software — he’s building a record. “California adopts responsibly, Washington blacklists arbitrarily” is a contrast he can run on.

The deal follows Newsom’s March executive order directing state agencies to develop new certification and procurement standards for AI vendors. That order required companies seeking government contracts to demonstrate responsible policies on bias, civil rights, and the prevention of misuse. The Anthropic partnership is the first major commercial agreement to emerge from that framework.

California built the regulatory framework first (March 2026), then validated it with a flagship deal second (June 2026). That’s the opposite sequence from the federal government’s approach, where export controls got issued in response to a specific jailbreak incident with no published standard for what triggers them.

The Sharpest Line in Today’s Coverage

One outlet’s framing captures the stakes more bluntly than any quote from either side: “The federal government put Anthropic on a blacklist. California put it in its government software portal. One of those decisions will look prescient; the other will look like a procurement mistake.”

That’s the bet being placed in real time, by two different layers of American government, about the same company.

Anthropic’s Strategic Calculation

For Anthropic, sitting in the middle of a federal export control negotiation, this deal serves multiple purposes simultaneously.

For Anthropic, the California deal is a commercial win at a moment when its enterprise strategy needs visible momentum. The company has committed 100 million dollars to its Claude Partner Network and is pushing Claude into large organisations through consulting firms like Accenture and Deloitte. California’s endorsement, coming from the state that hosts Anthropic’s headquarters, adds a prominent public-sector reference to that effort.

After eighteen days of headlines about export controls, jailbreaks, class-action lawsuits, and “they screwed us,” Anthropic needed a story that wasn’t about Washington. California gave them exactly that — a home-state endorsement at the moment they most needed one.

“As a California company, we feel a real responsibility to our home state. We’re honored to expand our partnership with California’s agencies and to put Claude to work for the people who keep this state running,” said Kate Jensen, Anthropic’s Head of Americas.

That “as a California company” framing is doing real work. It’s a quiet rebuttal to the federal “supply-chain risk” framing: we’re not a foreign threat, we’re your neighbor, headquartered in your state, run by your residents.

The Mythos Update Buried in the Coverage

One detail in the coverage that’s easy to miss but matters enormously: the government has since granted Anthropic approval to restore some access to its Mythos 5 AI model, according to a report last week from Bloomberg News.

The deal with California comes as Anthropic faces separate federal scrutiny over its powerful Mythos model, which was recently restricted over national security concerns before limited access was restored.

This confirms what we flagged yesterday (June 29) as “limited restoration” — now corroborated by Bloomberg reporting. The export control crisis is gradually, partially resolving, even as the broader Pentagon relationship remains adversarial enough that a federal judge had to intervene.

So the full picture as of June 30: Mythos partially restored. Pentagon designation judicially blocked. California signing the biggest state AI deal in US history with the same company. Three separate signals of de-escalation arriving within days of each other, even as the underlying relationship with the executive branch remains tense.

What This Means for the Broader Crisis Timeline

We’ve spent eighteen days documenting a single narrative: the US government versus Anthropic. Export controls. Jailbreak disputes. “They screwed us.” Staff flying to D.C. Class-action lawsuits. Austria recruiting Anthropic to Europe. China filling the gap with GLM-5.2.

Today complicates that narrative in an important way: “the US government” isn’t a monolith.

The federal executive branch — Commerce Department, Pentagon, Defense Secretary Hegseth — has treated Anthropic as an adversary needing control. The federal judiciary blocked at least one of those actions. And California, the fifth-largest economy on Earth with its own technology policy apparatus, just signed the largest state-level AI procurement deal in the country’s history with the exact company Washington spent eighteen days disabling.

This is federalism playing out in AI governance in real time. Different layers of American government, looking at the same risk profile, reaching opposite conclusions — and each layer has genuine authority to act on its conclusion. California can’t override an export control directive. But it can make Anthropic its default state AI vendor at scale, building a parallel track of legitimacy and revenue that doesn’t depend on Commerce Department goodwill.

My Take: The Reputational Rebuild Has a Home Field

Day 161, and I think this is the moment Anthropic needed since June 12, even if it doesn’t resolve anything structurally.

The export control crisis was never purely about a jailbreak. It revealed a relationship failure with the federal executive branch — a translation problem, as we called it on June 16, where Anthropic’s careful, evidence-based safety culture collided with an administration that rewards different signals entirely.

California offers Anthropic something the federal negotiation can’t: a government that already trusts them, says so publicly, and backs it with a 50% discount across every agency in the state. That’s not just commercial revenue. It’s a credibility anchor at a moment when credibility was Anthropic’s scarcest resource.

It also reveals something important about the limits of federal power over AI deployment. The Commerce Department can disable a model nationwide. It cannot stop a state government from publicly endorsing the company that built it, training thousands of employees on its tools, and building procurement infrastructure around it.

Last week, California launched a first-in-the-nation AI job-loss tracker designed to monitor whether the technology is displacing workers. California isn’t adopting AI uncritically — they’re building their own guardrail infrastructure in parallel. That’s the contrast Newsom is explicitly drawing: adoption with guardrails, not blacklisting without explanation.

Whether that contrast ages well depends entirely on what happens next. If Fable 5’s full restoration comes through smoothly and the Pentagon relationship thaws, California’s bet looks prescient — first mover on a partnership that paid off. If Anthropic’s models cause a genuine security incident down the line, the federal caution looks vindicated and California’s enthusiasm looks reckless.

One hundred sixty-one days into this crisis timeline, the most important AI governance story today isn’t about capability, benchmarks, or jailbreaks. It’s about which government you trust to assess risk correctly — and the answer, this week, is genuinely split between the people who run the Pentagon and the people who run Sacramento.

Two governments. One company. Diametrically opposite conclusions. Place your bets.

Keywords for June 30, 2026:

1. Federalism
2. Endorsement
3. Split
4. Legacy
5. Procurement